7.6: Inheritance from Keck HIRES/LRIS image storage format

The Keck HIRES and LRIS CCDs are often read out using 2 of the 4 amplifiers simultaneously. This briefly describes the conventions which are currently used for storage of image and calibration data from such readouts.

7.6.1: Layout of the Tektronix CCDs

Crudely, the format of the Tektronix 2k x 2k CCDs looks like this:
	|       !       |
	|               |   +-----> +X direction
	|       !       |   |
	|               |   |   Focal plane coords
	|       !       |   v
	|               |    +Y direction
	|       !       |

Here "<" and ">" designate readout amplifiers, and "!" designates
the boundary between the pixels read by the left and right amplifiers.
There are also amplifiers at the bottom corners, but in practice
a 4-amplifier readout is never done.  There is no keyword scheme
for describing a 4-amplifier readout.

(The orientation of X and Y here are chosen for conformance with video display devices. The diagram does not designate whether the illumination is coming from the front or back.)

7.6.2: Pixel storage schemes

In addition to the actual image pixels there may be overscan columns (OSC), prescan columns (PSC), overscan rows (OSR), and prescan rows (PSR) which contain data useful for calibration. Readout through two amplifiers introduces a complexity not found with single amplifier readouts.

For conceptual purposes only, the data from a dual amplifier CCD might be stored as seen below.

	| | PSR_1 | | | PSR_2 | |
	| |       | | |       | |       NOTICE:
	|P|       |O|O|       |P|
	|S|       |S|S|       |S|       This is not the scheme
	|C| Data1 |C|C| Data2 |C|       which is used.
	| |       | | |       | |
	|1|       |1|2|       |2|
	| |       | | |       | |
	| | OSR_1 | | | OSR_2 | |
This method of storage makes it quite clear which regions of data were read through which amplifier. Application of the calibration data could be done simply by splitting the image into 2 sections, using existing tools, and recombining the results. Unfortunately it also gives the impression that there is a spatial discontinuity down the middle of the CCD. The image data are completely contiguous at the middle. If the data were split in this fashion the image could not make use of a single set of WCS transformation keywords.

The Keck I instruments LRIS and HIRES have chosen a different scheme to handle the case where readout is done via more than one CCD amplifier. The pixel streams from the 2 CCD amplifiers are descrambled by the instrument computer. This provides a contiguous imaging region that can be described by a single WCS. Application of the calibration data requires a slightly more complicated version of cutting an pasting than above, but this is easy to hide in an IRAF procedure. In the descrambled memory the data are arrayed contiguously as indicated below. Each region of OSC and PSC consists of 2 halves from the 2 different amplifiers.

	| | | PSR_1 ! PSR_2 | | |
	| | |       !       | | |
	|P|P|               |O|O|
	|S|S|       !       |S|S|
	|C|C| Data1   Data2 |C|C|
	| | |       !       | | |
	|1|2|               |1|2|
	| | |       !       | | |
	| | | OSR_1 ! OSR_2 | | |
Note that the prescan row information from each amplifier is split into 2 non-contiguous sections. Likewise for the overscan row information from each amplifier.

PSC usually do not provide much calibration information and are often discarded entirely, but OSC are usually retained. Except during engineering tests PSR are never transmitted. OSR are commonly used by observers from Caltech but not by Lick observers.

7.6.3: FITS keywords documenting the readout

FITS keywords inserted in the headers document the number of pixels in each of the readout regions. They are not described here.

In normal operations at HIRES and LRIS only 2 amplifiers are used. A data storage scheme for all 4 amplifiers would require separation of OSR and PSR into separate chunks from the top and bottom amplifiers.

The CTIO ARCON detectors use a scheme similar to this.

Steve Allen <sla@ucolick.org>
$Date: 1996/03/18 20:27:24 $